Glasses capable of fusing at the lowest possible temperature to form a bond between the two basic parts of an electronic package (e.g. a die and its ceramic substrate) have long been sought and some success has been achieved. Exemplary of the current state of the art, in this respect, are the following:
______________________________________ U.S. Pat. No. 3,454,408 3,497,774 3,650,778 3,837,866 4,002,799 4,459,166 4,743,302 4,761,224 British Patent No. 1,552,648 ______________________________________
Representative of commercial materials embodying the disclosure of U.S. Pat. No. 4,459,166 are the die attach pastes known as JMI AuSub pastes sold by Johnson Matthy Corporation.
While some prior art references discuss the glasses employed in terms of their low melting temperatures, as recognized for example in U.S. Pat. No. 3,454,408, Col. 4, lines 24 et seq., it is perhaps more a-propriate to define such glasses in terms of their "glassy edge temperature" and for crystallinity characteristics in terms of their "devitrification edge". The disclosure herein adopts this terminology for purposes of defining and describing the subject invention. Thus the term "glassy edge temperature" is used herein to define that temperature at which the viscosity of the glass reaches about 10.sup.11.5 poises, thus permitting fusion of the glass particles (e.g. when in a paste) to take place. Likewise, the term "devitrification edge" is used herein to define a further point in the heating of the glass to fuse it (and thus usually a time-temperature relationship) before which (and thus usually a point of temperature below which) the glass after fusing at its glassy edge temperature begins to crystallize. The temperature difference between the "glassy edge temperature" and the "divitrification edge temperature" is referred to as the "glassy range".
Generally speaking, the above-recited prior art has had one or more limiting features imposed upon it. Firstly, to be truly suitable in the die attach area, most of the glasses are required to be heated to 400.degree. C. and often 450.degree. C. in order to obtain an acceptable bond of the die to the substrate. Such temperatures, while currently deemed "low", are still significantly high when it comes to handling integrated circuitry. Secondly, several of the known glasses are used in paste form with crystalline or silver flake additives, but which also use or require in addition to a solvent, a resin material for giving the paste the proper rheology for application. Such resins notoriously retain moisture, which has a significantly detrimental affect upon the life of the components in the package. Indeed, the military currently specifies in certain circumstances a limit of 5000 ppm on such moisture within the final hermetically sealed package.
Exemplary of one of the most efficacious of the prior art systems is the invention disclosed in U.S. Pat. No. 4,459,166 (and its parent, U.S. Pat. No. 4,401,768). While properly referring to itself as a low temperature system, the term is obviously relative. As reported in Col. 3, lines 4-6, the die attach temperatures contemplated are between 425.degree.-450.degree. C. A significant lowering of these temperatures would be a major advance in the art.
Yet another drawback to the development of acceptable low temperature glasses is the tendency of some glass, though having a very low glassy edge temperature, to crystallize excessively during the heating and cooing that is required to take place during formation of the bond during packaging. Stated another way, the glasses employed have too narrow a "glassy range", so that crystallization is virtually inevitable, absent very careful control. Such glasses are referred to as being "unstable", while those not having this problem are referred to as "stable" glasses.
Devitrification of the vitreous glass in a paste during bonding which has as its major goal the wetting of diverse electronic components, thereby to strongly adhere these components together, can be severely detrimental to the adhesive qualities of the ultimate bond structure, and can have other deleterious affects as well. This need to achieve a bond whose glass structure is substantially non-crystalline often eliminates numerous possible glass compositions that might otherwise have rather acceptable, low glassy edge temperatures.
To place the above in perspective, as far as die attaching is concerned, the loading of glasses with silver flake, called the Ag/glass method of die attach (reflected in some of the above references, e.g. see U.S. Pat. '774, '166, and '224 listed above) overcame several problems associated with the previously used Si/Au eutectic die attach methodology. Unfortunately, to date, this Ag/glass methodology, which would otherwise be extremely desirable to use, requires too high processing temperatures for the newer, higher density (submicron) semiconductor devices, silicon as well as gallium arsenide. Also, this higher processing temperature (400.degree. C.) creates an associated higher, undesirable stress on the device when cooled. Other disadvantages include:
1. The relatively high processing temperatures and times at temperature required to generate adhesion of the die to the package can also cause swimming of the lead frames in CerDip and oxidation of the gold-plated seal rings in multilayer ceramics The latter problem is the more serious when using gold-backed die, which generally requires higher processing temperatures. Nickel oxidation retards the wetting of the goldtin preforms used in the final seal operation and contributes to poor hermeticity yields. Nickel oxidation can be minimized by reducing the peak processing temperature, but normally at the sacrifice of adhesion, or at the very least, consistency of adhesion (reliability). To alleviate this compromise, the industry has largely been forced to use a forming gas clean-up cycle before final seal. This reduces the nickel oxide back to nickel and for the most part improves the gold-tin wetting to an acceptable level. However, complete wetting (and hermeticity) is not sufficient if excessive nickel has diffused through the gold. Increasing the gold-plating thickness and/or density retards the oxidation, but adds significant costs without completely solving the problem.
Thus, to process multilayer packages (PGA, CCC, Sidebraze, etc.), particularly with large area gold-backed die, requires not only undesirably high processing temperatures, but also necessitates three separate environments for processing--the initial firing in air, the oxide clean-up in forming gas, and the final seal in nitrogen, all normally in expensive conveyor furnaces.
2. The larger area die, normally associated with this package, requires excessively long times to drive the organics out without causing excessive voids. Die in excess of 400 mils on a side, largely common in gate array devices, can require a minimum of three hours drying. This results in an extremely low throughput and heavy capital investment. The "single pass" versions which reduce the drying time are largely unproven and create another set of problems.
It is apparent from the above that there exists a need in the art for a new glass composition, paste and method of use which overcomes the above-described problems. It is an object of this invention to fulfill this and other needs in the art which will become apparent to the skilled artisan once given the following disclosure.